[rfc-i] v3imp #9 Is "cbor" a type?

Julian Reschke julian.reschke at gmx.de
Fri Jan 23 03:43:42 PST 2015


On 2015-01-23 10:09, Sean Leonard wrote:
> Minor Improvement Need
> #9 Is "cbor" a type?
>
> This improvement questions the classification system of Section 2.49.5,
> the <sourcecode> @type.
>
> The type "cbor" probably refers to CBOR diagnostic notation. CBOR is a
> binary format, so <sourcecode> that is literally CBOR would have to show
> or represent actual octets. Compare "asn.1" vs. "ber/cer/der" (or "x.690").
>
> As with things like this, the problem is that now there is an informal
> registry that is supposed to be maintained by the RFC Editor. I am not
> in favor of a formal IANA registry. I am also not in favor of delegating
> the classification to media types, as many perfectly legitimate
> <sourcecode> types do not have media types (c++, c, perl, pseudocode,
> python, etc.). However we should recognize that "The RFC Editor will
> maintain a complete list of the preferred values on its web site, and
> that list is expected to be updated over time." is basically a registry
> by another name.
> ...

Indeed. My preference would be to use media types when they apply, and 
use short names otherwise.

> I suggest that @type be de-formalized, by permitting file extensions in
> widespread use. Thus "javascript" -> "js", "perl" -> "pl", "python" ->
> "py". One could add filenames to <sourcecode> elements (which was the
> original draft premise of Improvement #5); however, I think the point of

There is a @name attribute on sourcecode, no?

> <sourcecode> is for exemplary purposes with fragments (see Improvement
> #8). Therefore if file extensions are used, perhaps type=".js" (i.e.,
> prefix with a period) makes that really clear. I believe this has the
> right level of (in)formality, and doesn't get into esoteric things like
> "are we referring to Vernacular JavaScript 1.0 from the '90s or Strict
> ECMAScript 5".
>
> Of course the RFC Editor can still maintain lists of commonly used
> things, as well as things that are not file extensions per-se but still
> have identification value ("pseudocode").
>
> Oh yeah, "bash" -> ".sh". I found
> <http://unix.stackexchange.com/questions/31760/file-extensions-for-unix-shell-scripts>
> informative.

Best regards, Julian



More information about the rfc-interest mailing list